CN105099968A - Communication system at super-nyquist rate in multi-path channel - Google Patents

Abstract

The invention discloses a communication system at a super-nyquist rate in a multi-path channel. In the system, a sending end alternatively sends a useful signal and a training sequence; furthermore, a cyclic postfix is inserted; the signal is inserted into a cyclic prefix and transmitted into the multi-path channel after being subjected to super-nyquist mapping; channel estimation and equalization are carried out by utilizing a correspondingly received signal of the training sequence at a receiving end; the equalized signal is subjected to super-nyquist decision feedback equalization; intersymbol interference due to the super-nyquist effect is eliminated through iteration for many times; and the system performance is improved. The super-nyquist communication system disclosed by the invention is capable of carrying out signal transmission at the super-nyquist rate in the multi-path channel, so that the communication rate is increased.

Description

Super Nyquist rate communication system in a kind of multipath channel

Technical field

The present invention relates to communications industry technical field, particularly relate to the super Nyquist rate transmission systems in a kind of multipath channel.

Background technology

Super Nyquist sampling (Faster-Than-NyquistSignaling, FTN) theoretical, by corresponding even 40% the transmitting again without 10%, 20% of intersymbol interference transmission rate of the impulse waveform of the transmission rate of signal being brought up to transmitting filter, although introduce intersymbol interference artificially, but by suitable coding/decoding method, under signal transmitting power and the constant prerequisite of bandwidth, the progressive bit error rate performance of system can not decline.Although the implementation complexity of super Nyquist system is high, but along with the development of modern chips technology and the raising of computer computation ability, the software restraint of super Nyquist sampling system realizes becoming feasible.

And existing super Nyquist transmission system, when accelerated factor improves, when also namely super Nyquist speed improves, intersymbol interference more seriously causes systematic function to decline.

Summary of the invention

In order to overcome above-mentioned the deficiencies in the prior art, the present invention proposes the super Nyquist rate communication system in a kind of multipath channel.The super Nyquist rate transmission systems that the present invention proposes is applicable to the slow multipath channel become, and is not suitable for the fast channel become.

The intersymbol interference that the present invention is caused by decision feedback equalization algorithm elimination super Nyquist effect.On the other hand, channel estimation and equalization is carried out, the super Nyquist speed rates problem in solving multipath channel by sending training sequence.

For solving the problems of the technologies described above, technical scheme of the present invention is as follows:

A super Nyquist rate communication system in multipath channel, comprising:

At transmitting terminal, first carry out channel convolution code coding and interweaving encoding to data symbol, carry out phase-modulation, insertion training sequence and useful data are combined into Frame, after adding cyclic suffix, then carry out super Nyquist mapping.Super Nyquist signal is sent among multipath channel after adding Cyclic Prefix.

At receiving terminal, first corresponding according to the training sequence received signal, carries out the LS channel estimating based on DFT, according to the result of channel estimating, carries out MMSE channel equalization to super Nyquist signal.It is balanced that signal after channel equalization carries out decision-feedback super Nyquist again.Signal after equilibrium by log-likelihood ratio (log-likelihoodratio, LLR) demodulation, then carries out maximum a posteriori probability (Aposterioriprobability, APP) channel decoding, obtains decoded signal.

This system, by inserting training sequence, carries out estimating based on the least square (LeastSquares, LS) of discrete Fourier transform (DiscreteFourierTransform, DFT) to channel.According to the channel estimated, carry out equilibrium to received signal based on least mean-square error (Minimummeansquareerror, MMSE) criterion.Higher owing to working as super Nyquist transmission rate, that is to say when accelerated factor is less, the intersymbol interference that super Nyquist effect causes is also more serious, thus cause the decline of error rate of system performance, therefore by decision feedback equalization module, super Nyquist equilibrium is carried out to the signal after channel equalization, improve systematic function by the method for iteration.

Compared with prior art, the beneficial effect of technical solution of the present invention is: by decision feedback equalization algorithm, improves when accelerated factor increases, the performance of super Nyquist system; And system of the present invention employs the training sequence of excellent performance, do not need to estimate characteristic of channel adjusting training sequence in transmitting procedure.

Accompanying drawing explanation

Fig. 1 is workflow diagram of the present invention.

Decision feedback algorithms decoding performance schematic diagram when Fig. 2 is accelerated factor τ=1 (Nyquist rate).

The performance schematic diagram of the super Nyquist system that Fig. 3 is accelerated factor when being 0.8.

The performance schematic diagram of the super Nyquist system that Fig. 4 is accelerated factor when being 0.7.

Embodiment

Clear for making the object, technical solutions and advantages of the present invention estimate, below in conjunction with specific embodiment, and with reference to accompanying drawing, the present invention is described in more detail.Although herein can providing package containing the demonstration of the parameter of particular value, should be appreciated that, parameter without the need to definitely equaling the value responded, but can be similar to described value in acceptable error margin or design constraint.

Thought of the present invention is, by inserting training sequence, carries out channel estimating, and the channel according to estimating carries out channel equalization, then by decision feedback equalization algorithm, eliminates super Nyquist and transmit the intersymbol interference caused.

Specific implementation process of the present invention is:

First, every L data point of signal source forms a frame, then carries out chnnel coding.Chnnel coding uses constraint length to be 7, and coding polynomial is the convolution code of [171133].Wherein L=512, after chnnel coding, the data length of each frame is N=1024.

Sequence after coding carries out binary phase shift keying (BinaryPhaseShiftKeying, BPSK) modulation, and between each frame data after modulation, inserts training sequence.Training sequence uses the Chu sequence of length U=1024, and its expression formula is:

I road signal: $I\left(n\right)=cos\left({\mathrm{\&theta;}}_n\right)=cos\left(\frac{{\mathrm{\&pi;n}}^2}{U}\right),0\&le;n<U$

Q road signal: $Q\left(n\right)=sin\left({\mathrm{\&theta;}}_n\right)=sin\left(\frac{{\mathrm{\&pi;n}}^2}{U}\right),0\&le;n<U$

Again cyclic suffix is inserted to each frame signal.To fetch data front 2v data of frame, be inserted into the last of Frame.At receiving terminal, by removing the front v of receiving data frames and a rear v data, eliminate cyclic suffix.

Then super Nyquist mapping is carried out to data.

$s\left(t\right)=\underset{n=0}{\overset{N-1}{\&Sigma;}}a\left(n\right)p(t-n\mathrm{\&tau;}T)$

Wherein a (n) is for sending data symbol or Chu training sequence signal, and 0< τ <1 is accelerated factor, and Pulse shaped filter p (t) is square root raised cosine filter, and its expression formula is:

$p\left(t\right)=\frac{sin\&lsqb;\mathrm{\&pi;}(1-\mathrm{\&beta;})\left|\frac{t}{T}\right|\&rsqb;+4\mathrm{\&beta;}\left|\frac{t}{T}\right|cos\&lsqb;\mathrm{\&pi;}(1+\mathrm{\&beta;})\left|\frac{t}{T}\right|\&rsqb;}{\mathrm{\&pi;}\left|\frac{t}{T}\right|\&lsqb;1-{\left(4\mathrm{\&beta;}\frac{t}{T}\right)}^2\&rsqb;}$

Wherein rolloff-factor β=0.5, T=1, the auto-correlation function of transmitting filter p (t) is raised cosine filter, and namely its expression formula is:

$g\&lsqb;n\&rsqb;=\frac{\mathrm{\&pi;}}{4}\sin c\left(\frac{n}{T}\right)\&lsqb;\sin c(\mathrm{\&beta;}\frac{n}{T}+\frac{1}{2})+\sin c(\mathrm{\&beta;}\frac{n}{T}-\frac{1}{2})\&rsqb;$

Then get the rear portion of super Nyquist symbol, copy to front, also namely insert Cyclic Prefix.Circulating prefix-length is greater than multipath channel length, and desirable 10.

If the symbolic blocks length sent is N, circulating prefix-length 2v, then among Received signal strength, remove front v and a rear v symbol, then Received signal strength with the pass sending signal s be:

Wherein $\hat{y}={\&lsqb;y_0,y_1,...y_{N-1}\&rsqb;}^T,$ s＝[s ₀,s ₁,…s _N-1] ^T，

Tap coefficient matrix G ∈ R ^{n × N}, its expression formula is:

$G=\left[\begin{array}{ccccccc}g\left(-v\mathrm{\&tau;}\right)& g\left(-v\mathrm{\&tau;}+\mathrm{\&tau;}\right)& ...& g\left(v\mathrm{\&tau;}\right)& 0& ...& 0\\ 0& g\left(-v\mathrm{\&tau;}\right)& g\left(-v+\mathrm{\&tau;}\right)& ...& g\left(v\mathrm{\&tau;}\right)& ...& 0\\ 0& 0& g\left(-v\mathrm{\&tau;}\right)& ...& ...& ...& ...\\ ...& ...& ...& ...& ...& ...& ...\\ ...& ...& ...& ...& ...& ...& ...\\ ...& ...& ...& ...& ...& ...& ...\\ g\left(-v\mathrm{\&tau;}+\mathrm{\&tau;}\right)& g\left(-v\mathrm{\&tau;}+2\mathrm{\&tau;}\right)& ...& 0& 0& 0& g\left(-v\mathrm{\&tau;}\right)\end{array}\right]$

Multipath channel uses exponential damping channel.Send signal after multipath channel, received, the same to p (t) of matched filter expression formula by matched filter, it is combined response and carries out sampling and be g [n].

After carrying out discrete sampling by matched filter extraction signal, obtain the training sequence X sent _tScorresponding Received signal strength Y _tS, first estimate to obtain by LS to the estimated value of channel get IDFT, obtain:

$IDFT\{\hat{H}\&lsqb;k\&rsqb;\}=\hat{h}\&lsqb;n\&rsqb;=h\&lsqb;n\&rsqb;+z\&lsqb;n\&rsqb;,n=0,1,2,...N-1$

Order ${\hat{h}}_{DFT}\&lsqb;n\&rsqb;=\hat{h}\&lsqb;n\&rsqb;,0\&le;n<L;{\hat{h}}_{DFT}\&lsqb;n\&rsqb;=0,L<n<N.$ Wherein L is channel maximum delay.

Right make DFT, just can obtain LS channel estimation value H (z) based on DFT.

According to MMSE criterion, obtain channel equalization coefficients wherein for power and the ratio sending signal power of additive noise in channel.If the super Nyquist signal received is Y _fTN, then the super Nyquist signal after channel equalization is X _fTN=W (z) * Y _fTN.

Signal after channel equalization contains the intersymbol interference that super Nyquist accelerates to cause, and causes and eliminates this intersymbol interference by by decision feedback equalization.

Because G has loop structure, therefore can carry out Eigenvalues Decomposition to it, obtain: G=Q ^tΛ Q ^*

To received signal be multiplied by Q ^*transform to frequency domain, obtain:

$y_f\&cong;Q*\hat{y}=\mathrm{\&Lambda;}Q*s+Q*\mathrm{\&eta;}={\mathrm{\&Lambda;s}}_f+{\mathrm{\&eta;}}_f$

S _fand η _fthe signal of the corresponding frequency domain of difference and noise component(s).

MMSE frequency-domain equalizer W can obtain from eigenvalue matrix Λ.W is diagonal matrix, and the element of its i-th row i-th row is:

$w_{(i,i)}=\frac{{\mathrm{\&lambda;}}_{(i,i)}^{*}}{|{\mathrm{\&lambda;}}_{(i,i)}{|}^2+N_0}$

Wherein λ _{(i, i)}for the element of Λ i-th row i-th row, N ₀it is the power spectral density of noise.And then be multiplied by Q ^t, obtain the restituted signal estimated value of time domain:

Received signal strength through a super Nyquist frequency domain equalization, the frequency domain equalization result obtained then the Soft Inform ation fed back with channel decoder subtract each other, by this difference be input among DFF.The output of DFF is added with the output signal of frequency domain equalization, then carries out LLR solution mediation channel decoding, to improve decoding performance and operation efficiency.

Namely the input signal of LLR demodulator is:

The computing formula of LLR is:

Wherein σ ²for the variance of input data frame.The response of DFF is (B ^h-I _n), wherein matrix B can be released by super Nyquist interference matrix G.

If super Nyquist interference matrix is: G=[g ₀, g ₁, g ₂..., g _n-1] ^t

The then generation vector b of circular matrix B ₀meet:

$b_0={\&lsqb;1,b_1,b_2,\mathrm{...},b_{N-1}\&rsqb;}^T=\&lsqb;1,\frac{g_1^H{g}_0}{g_0^H{g}_0+{\mathrm{\&sigma;}}^2},\frac{g_2^H{g}_0}{g_0^H{g}_0+{\mathrm{\&sigma;}}^2},\mathrm{...},\frac{g_{N-1}^H{g}_0}{g_0^H{g}_0+{\mathrm{\&sigma;}}^2}\&rsqb;$

Decision-feedback matrix B is: B=[b ₀, b ₁, b ₂..., b _n-1]

Wherein vectorial b ₁, b ₂..., b _n-1by b ₀carry out cyclic shift respectively to obtain.Also be b ₁=[b ₁, b ₂..., b _n-1, 1] ^t, b ₂=[b ₂..., b _n-1, 1, b ₁] ^t, by that analogy.

The posterior probability information that channel decoding obtains, the input fed back by soft mapping block, the computing formula of Soft Inform ation is:

$\tilde{s}=\tanh \left(\frac{{\mathrm{APP}}^{\&prime;}}{2}\right)$

Wherein tanh () is hyperbolic tangent function, and APP ' is the posterior probability information that channel decoder exports.Fig. 2 and Fig. 3 be accelerated factor τ=0.8 and τ=0.7 time equilibrium result.After iteration twice, algorithm is restrained substantially.

Because the data of test are 10 ⁶bit, being therefore suitable in the error rate is 10 ^-3with 10 ^-4level on analyze simulation result.During accelerated factor τ=1, when namely sending data with Nyquist rate, owing to there is not the intersymbol interference that super Nyquist accelerates to cause, the result that therefore iterative algorithm is each is the same.And when accelerated factor τ=0.8, owing to there is the intersymbol interference that super Nyquist speed causes, after twice iteration, decision feedback equalization algorithm is just close to convergence.Bit error rate performance aspect, decision feedback equalization algorithm needs the noise Bizet of 4dB and 5.2dB to reach 10 respectively ^-3with 10 ^-4the error rate, close to result during Nyquist rate.

Obviously, the above embodiment of the present invention is only for example of the present invention is clearly described, and is not the restriction to embodiments of the present invention.For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description.Here exhaustive without the need to also giving all execution modes.All any amendments done within the spirit and principles in the present invention, equivalent to replace and improvement etc., within the protection range that all should be included in the claims in the present invention.

Claims (3)

1. the super Nyquist rate communication system in multipath channel, is characterized in that, comprising:

Dispensing device, for carrying out chnnel coding to transmission data, carrying out interweaving and modulate, and inserting training sequence, and interpolation cyclic suffix, carries out super Nyquist mapping by super Nyquist sampler to symbol, and add Cyclic Prefix and be transmitted among multipath channel;

Receiving system, by carrying out channel estimating and channel equalization to received signal, by decision-feedback iteration equalizing algorithm, eliminating super Nyquist and mapping the intersymbol interference caused; The reception data that receiving system is first corresponding according to training sequence, carry out channel estimating; According to the estimated value of channel, channel equalization is carried out to reception data symbol; Again by super Nyquist DFF, eliminate the intersymbol interference caused by super Nyquist speed rates.

2. the super Nyquist rate communication system in multipath channel according to claim 1, is characterized in that, every N number of data point of signal source is formed a frame by transmitting terminal, then carries out chnnel coding; Sequence after coding carries out binary phase shift keying BPSK modulation, and between each frame data after modulation, inserts training sequence; Insert cyclic suffix to each frame signal again, super Nyquist mapping equation is:

$s\left(t\right)=\underset{n=0}{\overset{N-1}{\&Sigma;}}a\left(n\right)p(t-n\mathrm{\&tau;}T)$

Wherein a (n) is for sending data symbol or training sequence signal, and 0< τ <1 is accelerated factor, and N represents the length sending Frame; Pulse shaped filter p (t) is square root raised cosine filter, and its expression formula is:

$p\left(t\right)=\frac{sin\&lsqb;\mathrm{\&pi;}(1-\mathrm{\&beta;})\left|\frac{t}{T}\right|\&rsqb;+4\mathrm{\&beta;}\left|\frac{t}{T}\right|cos\&lsqb;\mathrm{\&pi;}(1+\mathrm{\&beta;})\left|\frac{t}{T}\right|\&rsqb;}{\mathrm{\&pi;}\left|\frac{t}{T}\right|\&lsqb;1-{\left(4\mathrm{\&beta;}\frac{t}{T}\right)}^2\&rsqb;}$

Wherein rolloff-factor β=0.5, T=1.

3. the super Nyquist rate communication system in multipath channel according to claim 1, is characterized in that, first receiving terminal carries out channel estimation and equalization, then carries out super Nyquist decision feedback equalization; After carrying out discrete sampling by matched filter extraction signal, obtain the training sequence X sent _tScorresponding Received signal strength Y _tS, first by obtaining the estimated value H (z) of channel based on the least square LS method of estimation of discrete Fourier transform DFT;

According to least mean-square error MMSE criterion, obtain channel equalization coefficients:

$W\left(z\right)=\frac{H^{*}\left(Z\right)}{|H\left(z\right)|2+{\mathrm{\&sigma;}}_n^2/P}$

Wherein for power and the ratio sending signal power of additive noise in channel; If the super Nyquist signal received is Y _fTN, then the super Nyquist signal after channel equalization is X _fTN=W (z) * Y _fTN;

Signal after channel equalization contains the intersymbol interference that super Nyquist accelerates to cause, and causes and eliminates this intersymbol interference by by decision feedback equalization.